Lucas (TaeYoung) Jo

Contacts

Professional Summary

Architecting the Future of Voice Interface: Bridging Hardware-Level DSP with Agentic AI Orchestration.
A visionary technology leader and R&D architect with 16+ years of multi-disciplinary expertise spanning from Silicon-level Audio Engineering to Large-scale Cloud AI Systems.

As the CTO of Atlas Labs (and Board Director at the US parent company, Atlasguide), I architected and scaled a 1M-user AI-telephony infrastructure (Switch), pioneering the integration of LLM-based agents with real-time VoIP stacks. I am the co-creator of the Zeroth Project, which established the official Korean ASR baseline in the global Kaldi repository (SLR40)—now an industry standard for Korean speech research. My current work at JW.org World Headquarters (NY, USA) involves leading AI initiatives for low-resource languages, focusing on multilingual ASR improvements and scalable research infrastructure. Details are omitted to respect confidentiality.

I am uniquely positioned at the intersection of (1) low-level audio signal processing, (2) high-availability cloud infrastructure, and (3) agentic AI orchestration, with a proven track record of solving high-complexity communication challenges at a global scale.

Project Portfolio (2017–Present)

Focus: National-scale voice infrastructure, enterprise deployments, and humanitarian multilingual access.

Zeroth Open Source ASR (2017–2018) — OpenSLR SLR40 Korean baseline, 357 GitHub stars, 120 forks.
B2B ASR Market Development (2018–2019) — Enterprise adoption: Sorijava, PoscoICT, Mindwareworks, Yes24, Saltlux.
Zeroth Decoder (2020–2022) — Production-grade streaming ASR engine, v2.5.8–v2.7.0 releases.
Switch AI‑Telephony Platform (2019–2024) — 30M+ calls, 10K+ daily calls, 119 microservices on AWS EKS.
Japan Market Expansion (2020–2022) — FreeBit/IPSPRO integration, 500% growth in JP traffic.
Sentroid Contact Center Analytics (2022–2024) — Genesys Cloud integration; enterprise customers Coupang, Amore Pacific, KCT.
Rachel AI Voice Agent (2023–Present) — Real‑time voice AI, Go backend, 50+ commits.
Acomm AI Platform (2025–Present) — Hexagonal architecture; Vision‑first RAG; multimodal workflow engine.
AI Receptionist “Leah” (2024–Present) — U.S. healthcare‑oriented telephony integration (Twilio/Daily).
JW.org OmniASR (2024–Present) — Humanitarian project for 1,000+ low‑resource languages.

Skills

Conversational AI & Agents: Pipecat, LangGraph, MCP (Model Context Protocol), RAG, LLM Orchestration, Real-time Voice Agents (Project Leah).
ASR & Speech AI: Meta Omni ASR (7B ZS), Few-shot Learning (ICL), Whisper (Fine-tuning), Kaldi (Zeroth Project/SLR40), Wav2Vec 2.0, ONNX Optimization, Nvidia Triton, TensorRT-LLM.
Audio Engineering: Noise Suppression (Resemble Enhance), Audio Effects (FFmpeg, NumPy-based DSP), OpenVoice, Serverless Audio APIs, Class-D PWM Modulator Design.
VoIP & Infrastructure: Asterisk (ARI), Kamailio, SIP/WebRTC (Janus), AWS (EKS, Lambda Layer Optimization), S3-native Training, WebDataset, ArgoCD (GitOps), Terraform (IaC), RunPod/GPU Automation, Zero-Downtime Cluster Switching.
Leadership & Strategy: CTO Leadership (Board Member of US Parent), Startup Co-founding, R&D Management, Global Technical Advisory, Industrial Standardization (SLR40).

Professional Experience

Atlas Labs / Atlasguide | Jan 2017 – Present

CTO / Head of Technology (Jan 2023 – Present)

Executive Leadership: Serving as the CTO and Board Director at the US parent company, Atlasguide. Driving technical strategy and long-term R&D roadmaps for AI-driven communication products.
Agent-VoIP Orchestration: Architected a seamless bridge between LLM agents and VoIP stacks, treating the telephony layer as a programmable “communication tool.”
Project Leah (AI Agent System): Directed the R&D of a bidirectional voice agent system. Deployed RAG-enabled agents for inbound reception and autonomous outbound booking with real-time language detection. Built using Pipecat, orchestrated via LangGraph and MCP.
Engineering Excellence: Designed a high-velocity delivery pipeline for 10+ microservices using a Monorepo, ArgoCD, and GitHub Actions. Established a Terraform foundation enabling Zero-Downtime Cluster Switching (99.9% availability).
Operational Efficiency: Automated 90% of infrastructure management, effectively eliminating the need for a 2-person DevOps team while improving productivity via Tailscale-based local-to-cloud debugging.
Whisper Optimization: Fine-tuned Whisper-large v3 using thousands of hours of call logs. Built a high-performance gRPC inference server using Nvidia Triton and TensorRT-LLM.

VoIP Infrastructure Team Lead (Jan 2020 – Dec 2022)

Switch App Success: Built and launched the Switch app (1M+ users, 10,000+ paid subscribers). Managed infrastructure processing 15,000+ hours of calls monthly.
STT Accuracy Breakthrough: Enhanced call-domain CER from 75% to 89% for Korean and Japanese using Meta’s Wav2Vec and unsupervised pre-training.

Machine Learning Team Lead (Jan 2017 – Dec 2019)

Zeroth Project (SLR40): Led the development of the Zeroth Project, establishing the first official Korean recipe in the Kaldi repository.
Collaborative R&D: Led the Language Model (LM) development (text normalization, morpheme analysis, pronunciation rules), creating a foundation that democratized Korean ASR for researchers globally.
Commercialization: Developed a C++ multithreaded Kaldi decoder for real-time commercial use, bridging the gap between research and production.

JW.org (WHQ MEMPS, New York, USA) | Jan 2024 – Present

Collaborating with the World Headquarters (WHQ) to solve high-complexity language challenges for global non-profit initiatives. Public details are intentionally limited to honor confidentiality and IP ownership requirements.

Micro Team Lead, AI Services (Jul 2025 – Present)

Multilingual ASR Initiatives: Led low-resource language projects, improving accuracy through model adaptation and evaluation.
Scalable Data Pipeline: Built scalable data pipelines for large audio corpora and research workflows.
R&D Automation: Automated research environment provisioning to improve reproducibility and setup speed.

AI / Audio Engineer (Jan 2024 – Jun 2025)

Serverless Audio Engine: Architected serverless audio processing for real-time enhancement with latency and footprint optimization.
Runtime Optimization: Reduced runtime dependencies and optimized packaging for serverless deployment reliability.
Noise Suppression: Integrated noise suppression workflows and tuned environment presets for low-latency inference.

Foundations & Hardware Innovation

RADSONE | Senior Research Engineer | Jun 2015 – Sep 2016

Ripplebuds Project: Analyzed the “occlusion effect” in-ear microphones; developed a restoration system using a Neural Network-trained multi-band equalizer.
airDAC Project: Designed high-end Wi-Fi/USB DAC and headphone amplifier circuits; exhibited at the 2016 Munich High-end Show with critical acclaim.

EXTEGER | Co-founder / Senior Research Engineer | May 2014 – May 2015

Fabless Semiconductor Startup: Co-founded a startup specializing in ASIC design for Class D amplifier PWM modulators; successfully supplied chips to Samsung and LG.
Algorithm Design: Developed a click-free PWM switching frequency shift algorithm and an envelope-based PVDD control system, achieving 40% efficiency improvement.

PULSUS | Research Engineer | Mar 2010 – Apr 2014

ASIC Front-end Design (PS8645): Developed an integrated one-chip solution (USB/PWM/Class-D) adopted by SONY for iPhone docking speakers.
Digital Audio Engine: Engineered a high-performance PWM engine using MATLAB simulations, improving THD by 20dB and achieving SNR > 115dB.
AATP Platform: Developed a real-time audio algorithm verification tool (C++/Qt) as a Winamp plugin for rapid DSP prototyping.

LG Telecom | Assistant Research Engineer | Feb 2007 – Aug 2008

Optimized EVDO rev.A wireless network coverage and throughput.

Public Speaking & Technical Education

Hanyang University (ASML Lab) Invited Speaker
- “Kaldi based ASR system” (Aug 2019): Technical seminar on building production-grade ASR systems for graduate researchers.
- “Data-driven Way of Korean LM Design” (Feb 2018): Methodology for optimizing Korean Language Models.
FastCampus Lead Instructor & Subject Matter Expert
- Developed and led two consecutive intensive professional courses (30 hours each, 10 sessions x 3 hours) on Deep Learning-based Speech Recognition and Kaldi System Implementation.
K-Mobile Academy Invited Instructor
- “Understanding Speech Recognition & Kaldi System Implementation”: 1-day intensive workshop for industry professionals.

Education & Key Achievements

Master of Science in Electrical and Computer Engineering, Seoul National University (2007)
Bachelor of Science in Electrical and Computer Engineering, Seoul National University (2005)
Industrial Standard: Co-creator of Kaldi SLR40 (The official Korean ASR baseline).
Startup Scale: CTO of Atlas Labs (4B+ KRW investment, 1M users).
Global Impact: Leading “AI for Social Good” initiatives at JW.org WHQ for minority languages.